MCP Registry

@hazeljs/mcp

91bd98b6-d1a3-5b49-8dcc-4f478917403b | Store v2.1.3
31/ 100
LOW risk
No change since v2.1.2
AI review verdict
No high-risk signal observed

Based on the RiskyPlugins AI security review of the observed evidence.

No individual score drivers were recorded for this analysis.

Analysis record

Analysed
Today 22 November 2026
Version
v2.1.3
Artifact
SHA256 3cbd8766…f8b8ec
Detail
Full list

Evidence ledger

Ranked by severity · findings with a source location link to the code viewer

1 finding

Finding Categories

1
Network

MCP Server Analysis

MCP servers expose tools and resources to AI assistants. Unlike browser extensions, they run as standalone processes with direct system access. Tool definitions are analyzed for prompt injection, data exfiltration, and tool poisoning patterns.

AI Security Report

AI Security Review

Evidence context: threat category none; evidence quality strong.

Tool-Poisoning Assessment

The evidence shows zero tool-poisoning findings (finding summary: "tool-poisoning":"0"). This is significant because tool poisoning is the defining threat for MCP servers. The package description states it "Expose HazelJS tools as MCP tools over STDIO," which is a legitimate adapter pattern. There are no findings of hidden AI directives, invisible Unicode characters, or XML-style instruction tags in tool descriptions.

Credential and Network Access

The evidence shows zero secret/credential findings ("secret":"0"). The package does not read sensitive paths like .ssh/, .aws/, .kube/, or target specific environment variables like GITHUB_TOKEN or AWS_SECRET_ACCESS_KEY. There is only one network finding: NET-FETCH-dist/client/mcp-client.js-159, which is a standard fetch call in an MCP client implementation. This is expected behavior for an MCP server that communicates over STDIO transport.

IOC Findings Analysis

The 60 IOC findings are XIOC extractor false positives:

  • XIOC-DOMAIN-hazeltooladapter.js.map - a source map file path, not a domain
  • XIOC-DOMAIN-object.prototype.hasownproperty.call - JavaScript property access chain, not a domain
  • XIOC-DOMAIN-request.id - JavaScript property access, not a domain
  • XIOC-DOMAIN-param.name - JavaScript property access, not a domain
  • XIOC-URL-https://img.shields.io/ - legitimate npm badge URLs for documentation
  • XIOC-DOMAIN-www.npmjs.com - legitimate npm registry

These match the documented XIOC false-positive patterns: property access chains misread as domains and source map files flagged as domains.

Malware-Signature Findings

The 38 high-severity malware-signature findings are YARA code-smell rules firing on bundled/minified JavaScript in the dist/ directory. With zero actual malware findings ("malware":"0"), these are expected false positives from bundled dependencies.

Strongest Counterargument

The high finding count (103 total) and 38 high-severity malware-signature findings could suggest malicious code. However, the absence of tool-poisoning, credential access, and suspicious network destinations (only npm badges and legitimate MCP client code) indicates these are false positives from bundled dependencies and XIOC extractor noise. The package's stated purpose as an MCP adapter aligns with its minimal, legitimate code behavior.

Key Reasons

  • Zero tool-poisoning findings - the defining MCP threat is absent
  • Zero credential/secret access findings - no sensitive path reads
  • All IOC findings are XIOC false positives (source maps, property chains, npm badges)
  • Single network call is legitimate MCP client fetch in dist/client/mcp-client.js
  • Malware-signature findings have zero actual malware matches - YARA code-smell noise

False Positive Considerations

  • XIOC extractor misreading source map files as domains (.js.map)
  • XIOC extractor misreading property access chains as domains (object.prototype.hasownproperty.call)
  • YARA code-smell rules on bundled dist/ directory
  • NPM badge URLs flagged as suspicious IOCs

Reviewed 2026-04-27; recommended action: suppress false positive; model confidence 85%.

MCP version history

Risk trend by version

27 analyzed versions. Each point is the latest successful scan for that version; failed zero-score scans are hidden. Dates: First seen by Risky Plugins.

Selected
31
Change since first
-4
Change from previous
No change
Versions:
First analyzed version
0.8.6
First seen May 9, 2026
Risk range
31 to 43
Across analyzed versions
Latest analyzed version
2.1.3
First seen Nov 22, 2026
Selected version
low
Version
v2.1.3
First seen Today
Risk score
31
Findings
1
Change vs previous
No change

Pick any point on the chart, then open the Source tab to read that version's code.

Source Code Viewer

Source Code Not Available

Source code is not available for this version of the extension.

About This Extension

Expose HazelJS tools as MCP (Model Context Protocol) tools over STDIO

Frequently Asked Questions